Hostkey NL oVirt VPS — 8 vCore / 32 GB RAM / 2×4 TB HDD (one ext4/LVM root disk, one standalone local-zfs pool). Runs 4 production VMs.
| VM | Name | vCPU | RAM | Role |
|---|---|---|---|---|
| 100 | sn-infra | 1 | 4 GB | Forgejo, PowerDNS-Admin, Namevault, this wiki |
| 102 | sn-web | 2 | 4 GB | Stratus Digital's client sites |
| 104 | sn-monitor | 1 | 4 GB | Grafana/Prometheus/Loki/Uptime Kuma |
| 106 | sn-security | 2 | 8 GB | Wazuh SIEM (cores capped at 2 permanently — see Operational Rules) |
Before adding a new VM or service, check RAM/disk headroom first — zfs list local-zfs / free -h on the target host, not assumed.
4 vCPU / 8 GB RAM / 120 GB NVMe. Edge only — Caddy, DNS, mail, WireGuard, CrowdSec. Never deploy databases or media here (Vaultwarden is the one deliberate exception, kept here specifically so credentials survive a bare-metal outage).
Standalone Hostkey VPSes, not Proxmox VMs:
vm.v2-medium — 8 vCPU / 16 GB RAM / 160 GB NVMe (Hostkey CH)vm.v2-nano — 2 vCPU / 4 GB RAM / 60 GB NVMe (Hostkey NL)Local hardware (not a rented VPS), two disks: sda (5.5 TB, OS) and sdb (3.6 TB, the actual backup datastore — currently the single point of failure for anything backed up there, no RAID).
The original AMD EPYC 3151 host is permanently gone (unfixable NIC fault, refunded). If HoT ever colocates real hardware again: that CPU generation had no AVX-512, requiring cpuv1 image tags for affected workloads (MinIO/Notesnook was the known case). Not relevant to the current fleet, which runs generic virtualized Intel cores.